Bug 1244: Added description for pospopcount
authorShriya Sharma <shriya@redsemiconductor.com>
Tue, 9 Jan 2024 13:32:09 +0000 (13:32 +0000)
committerShriya Sharma <shriya@redsemiconductor.com>
Tue, 9 Jan 2024 13:32:09 +0000 (13:32 +0000)
conferences/fosdem2024/fosdem2024_ddffirst/fosdem2024_ddffirst.tex

index ea65ac8ebe2dfd3e098ef1096e397844337446cb..0f42add4bccd30b910b6cb491797bae58181fa55 100644 (file)
@@ -170,7 +170,7 @@ for (i = 0; i < VL; i++)
                identical to general-purpose Simple-V DD-FFirst...
        \end{itemize}
        
-}
+}Po
 
 \frame{\frametitle{maxloc}
   \begin{itemize}
@@ -183,6 +183,7 @@ for (i = 0; i < VL; i++)
   \begin{itemize}
        \item   Positional popcount adds up the totals of each bit set to 1 in each bit-position, of an array of input values.
        \item   Notoriously difficult to do in SIMD assembler: typically 550 lines
+
    \end{itemize}
        
        \lstinputlisting[language={}]{pospopcount.c}
@@ -192,7 +193,7 @@ for (i = 0; i < VL; i++)
 \frame{\frametitle{Pospopcount}
        
        \begin{center}
-               \includegraphics[width=0.6\textwidth]{pospopcount.png}
+               \includegraphics[width=0.5\textwidth]{pospopcount.png}
        \end{center}
        
 }
@@ -200,8 +201,17 @@ for (i = 0; i < VL; i++)
 \frame{\frametitle{Pospopcount}
        
        \begin{center}
-               \includegraphics[width=0.6\textwidth]{array_popcnt.png}
+               \includegraphics[width=0.5\textwidth]{array_popcnt.png}
        \end{center}
+
+  \begin{itemize}
+               \item   Part of the challenge is therefore to perform an appropriate transpose of the data,
+                               in blocks that suit the processor and the ISA capacity.
+               \item   The gbbd instruction is used for implementing the transpose function, 
+                               preparing the data for using the standard popcount instruction.
+
+       
+       \end{itemize}
        
 }